Introduction

Elemental phosphorus (P) is not only ubiquitous in human life, it is also one of the most fascinating areas of chemistry as it can exist in a large diversity of allotropes1,2,3, in various cluster configurations4,5, or in organic compounds6. The phosphorous polymorphism is even multiplied on the atomic scale when using a surface to constrain the reaction in two dimensions (2D). As in the field of on-surface chemistry producing complex nanographene structures in ultra-high vacuum (UHV)7,8, surface-assisted phosphorus reactions on metals have synthesized blue phosphorus9, P chains10, or even planar cyclo-P5 rings11,12. Since then, phosphorus allotropes have emerged as a promising one-atom thick 2D material beyond graphene, due to its moderate direct band gap (0.3 to 2.0 eV)13 suitable for nanoelectronics and nanophotonics applications14,15. However, allotropic configurations, their atomic buckling, defects, or potential alloy formation can be detrimental to the semiconducting character. In addition, the interaction of 2D materials with delocalized electrons of metal, as well as the dynamical charge transfer between the two media, are key factors in fostering gate-tunable functionalities such as superconductivity16,17. Experimental study of these aspects at the fundamental level is therefore essential for future quantum applications where metallic electrical contacts are required18.

Low-temperature scanning probe microscopy is an incontrovertible tool for assessing atomic structures in contact with metals and characterizing their electronic properties with high spectral resolution in UHV. Atomic force microscopy (AFM) with functionalized tips19,20 has demonstrated real-space imaging with improved lateral and vertical resolution of aromatic molecules and cyclo-carbons21. Recently, AFM imaging and spectroscopy have also tackled monoelemental 2D materials demonstrating a precise quantification of the atomic buckling in these structures22,23. Not only restricted to structural characterization, charge distributions, and work function changes are also accessible at the nanometer scale using Kelvin probe force microscopy (KPFM)24,25,26,27. In addition, the investigation of the local density of states (LDOS) of 2D materials near the Fermi level is readily achieved by means of STM and scanning tunneling spectroscopy (STS). Tunneling spectroscopy can also probe the image potential states (IPS) of 2D synthetic materials, as demonstrated in the case of graphene28, germanene29, or borophene30,31. Quantifying these Stark-shifted unoccupied states lying below the vacuum level gives not only access to the fundamental physical processes involved in charge carrier dynamics but also to quantify local modulations of the work function at the interface between 2D materials and metals.

By applying this methodology, we determine here the structure of phosphorus chains and self-assembled cyclo-P5 pentamers on Ag(111) using low-temperature (4.5 K) AFM imaging with CO-terminated tips. KPFM spectroscopic measurements indicate the formation of an inwards dipole moment at the P5/Ag interface, which results from the charge transfer from the Ag substrate to the network, as confirmed by DFT calculations. This charge transfer leads to a complex charge redistribution and the formation of an interfacial hybridized state (IS). Through FERT and STS spectroscopy, we determined the energy position of the IS and the series of IPS at the cyclo-P5 assembly as compared to pristine Ag, confirming an increase of the local work function of  ≈ 0.46 eV. We found that the P5/Ag system behaves as a prototypical p-type semiconductor-metal junction with a Schottky barrier built at the interface, opening prospects for its use in field-effect transistors, diodes, or solar cells. Given the general interest in tailoring the physical characteristics of monoelemental 2D materials contacted to a metal, we think that our experimental approach might serve as a powerful asset for deciphering the energy level alignment of future devices involving electric contacts.

Results

Atomic-scale imaging of the cyclo-P5

Phosphorus atoms were sublimed in UHV onto the Ag(111) substrate, kept at about 423 K (see “Methods”). Figure 1a shows an STM overview image of the resulting structures for a relative coverage of less than 0.3 monolayer (ML). Similar to refs. 10,11, extended 1D chains aligned along the \([1\overline{1}0]\) directions of Ag(111) (marked 1) in Fig. 1a) coexist with domains of P5 molecules (2). The inset of Fig. 1a further shows a STM image of the double and triple chains, which depends on the P deposition rate11.

Fig. 1: Atomic structure of phosphorus chains on Ag(111).
figure 1

a STM topographic image after sublimation of phosphorus atoms on Ag(111) leading to P chains (1) and cyclo-P5 domains (2), (It = 1 pA, Vs = 0.15 mV). The inset shows an STM image of the single, double, and triple chains, respectively. b–d Series of AFM images with CO-terminated tip revealing the armchair structure of single, double, and triple P chains, (f0 = 26 kHz, A = 50 pm). Scale bars are 0.5 nm. e Atomic configurations of the triple armchair chains obtained by DFT calculations. Phosphorus and silver atoms are shown in orang and gray, respectively. f Corresponding AFM simulation using the DFT coordinates.

Full size image

To precisely determine their atomic configurations, we employed AFM imaging with CO-terminated tips (see “Methods”, Fig. 1c, d)19. A common AFM contrast is observed for all configurations assigned to an armchair structure of the chain, which resembles that of hydrocarbon chains32. In each configuration, the chain has always an apparent width by AFM of 6.0 Å with a minimum distance between them of 0.25 Å. Thus, double and triple chains have apparent widths of 14.5 and 23 Å, respectively. The relaxed structure of the triple chain configuration calculated by DFT is shown in Fig. 1e. Phosphorus atoms colored in orange sit on bridge sites of the Ag lattice (gray) and are aligned along one \([1\bar{1}0]\) direction in accordance with the experimental data. Based on the DFT coordinates, we simulated the AFM image (see “Methods”, Fig. 1f). The excellent agreement with the experimental image of Fig. 1d confirms the armchair structure of the P chains on Ag(111), similar to ref. 10.

Increasing the P coverage to about 0.4 ML while keeping the substrate at 423 K leads to the formation of large islands of cyclo-P5 pentamers relative to the chains (Fig. 2a)12. In Fig. 2b, the close-up STM image reveals the structure of the self-assembled domains consisting of a hexagonal lattice with parameters a1 = b1 = 7.6 Å. Each bright protrusion corresponds to one cyclo-P5 molecule as schematized by the black dashed pentagons. Domains of cyclo-P5 rings exhibit a superstructure characterized by stripes separated by  ≈ 4.56 nm (i.e., six P5 rows), as shown by black dotted lines in Fig. 2a. These lines are rotated by 19° as compared to the \([1\bar{1}0]\) directions of the Ag(111) substrate, which agrees with previous experimental works11 as well as the relaxed structure obtained by DFT calculations of Fig. 2d. We thus confirm that the cyclo-P5 assembly adopts a (\(\sqrt{7}\times 6\sqrt{7}\))R19° unit cell with respect to the Ag(111) surface lattice as previously reported by Zhang et al.11.

Fig. 2: Atomic structure of the self-assembled cyclo-P5 molecules.
figure 2

a STM image of the self-assembled pentamers on Ag(111), (It = 1 pA, Vs = 0.15 mV). Islands systematically show a superlattice of bright lines rotated by 19∘ with respect to the \([1\overline{1}0]\) directions of Ag(111) (dotted lines). b Close-up STM topography showing the P5 pentamers depicted by dashed lines. c Corresponding AFM image revealing the P5 chemical structure, (f0 = 26 kHz, A = 50 pm). Each cyclo-P5 ring is composed of three short bonds (blue) and two longer ones (red), respectively. d Atomic configurations of the pentamer assembly on Ag(111) obtained by DFT. Phosphorus and silver atoms are shown in orange and gray, respectively. e Corresponding AFM simulation using the DFT coordinates. f Site-dependent Δf(Z) spectroscopic curves acquired at two neighboring P atoms of a P5 molecule (orange and brown), between two P5 molecules (gray) and on Ag(111) (black), respectively. The local minima of the Δf(Z) curves indicate the relative height of the phosphorus atoms.

Full size image

A deeper insight into the chemical structure of the cyclo-P5 molecules is provided by the AFM image in Fig. 2c. We also simulated the AFM image based on DFT coordinates, allowing us to confirm the exact position and structure of the P5 molecules in their self-assembly in registry with the Ag(111). The P-P bond length within the cyclo-P5 pentamer extracted by AFM varies from 2.5 to 2.7 Å, which is always larger by about 8% than that of DFT calculations for the pentamer on Ag(111) (Fig. 2d and Supplementary Fig. 1) or in the gas phase (2.185 Å)33. It is well-established that this overestimation of apparent bond lengths is induced by the tilting of the CO molecule attached to the AFM tip upon scanning, as shown for planar polyaromatic hydrocarbons or P3N3 molecules34,35. Moreover, each cyclo-P5 ring is composed of three short apparent bonds (2.5 Å, blue bonds in Fig. 2c) and two longer ones (2.7 Å, red bonds). This particular bond order, also confirmed in the relaxed structure calculated by DFT (Fig. 2d), is likely induced by the small buckling of the structure when adsorbed on Ag(111).

To accurately quantify the atomic corrugation within the cyclo-P5 structure, we acquired a series of site-dependent Δf(Z) spectroscopic curves (Fig. 2f) at the locations marked in the inset AFM image. The black and gray curves were obtained on Ag and between two pentamers, respectively. On top of neighboring atoms of a cyclo-P5 (orange and brown curves), spectra exhibit a characteristic dip arising from the interaction between the front-end oxygen atom of the CO-terminated tip with the phosphorus atom. The dashed vertical lines indicate the Z position of their bottoms and are the signature of the relative atomic Z height23. The difference ΔZ of  ≈ 20–30 pm thus represents the intrinsic atomic corrugation within the cyclo-P5 pentagonal structure5, which is comparable with atomic corrugations in graphene36 or planar molecules37. Thus, this confirms the planarity of the cyclo-P5 structure5, as reflected in the constant-height AFM image of Fig. 2e.

Charge distribution at the P5/Ag interface

Through DFT calculations, it has been determined that the cyclic P5 pentamer exhibits a higher binding energy on the Ag(111) surface, amounting to − 0.90 eV per atom (Supplementary Fig. 2). This larger energy value is primarily facilitated through a charge transfer mechanism, promoting the stability of the cyclic P5 structure. To provide insights into the charge distribution at the cyclo-P5 interface, we performed force versus voltage spectroscopic measurements (see “Methods”). Experimentally, the frequency shift Δf as a function of the sample bias Vs is measured at a constant tip height Z, providing in the Δf(V) curve a parabola due to the electric force acting between tip and sample. The voltage V* at the top of the parabola represents the local contact potential difference (LCPD) between tip and sample, which allows one to image charge distributions and work function changes with nanoscale resolution24,25,26,27. Figure 3a shows a Δf(V) cross-section acquired across a P5 domain (see STM inset of Fig. 3a). Single Δf(V) point-spectra on top of the P5 network (orange) and on Ag(111) (black) are plotted in Fig. 3b, respectively. The dashed lines in Fig. 3a, b refer to the V* position. The LCPD value systematically shifts towards positive values (ΔV* ≈ 0.22 V) for the pentamer assembly as compared to the pristine Ag substrate. This indicates the accumulation of charges at the P5 network as compared to the Ag substrate.

Fig. 3: Charge redistribution at the cyclo-P5/Ag(111) interface.
figure 3

a Frequency shift Δf as a function of sample bias voltage Vs, measured across a pentamer domain shown in the STM image by a dashed line, (parameters : f0 = 26 kHz, A = 80 pm). b Single Δf(V) curves at the pentamer assembly (orange) as compared to the Ag(111) (black). Dashed lines mark the top of the parabola allowing the extract an LCPD shift ΔV* = 0.22 V. c Top and side views of the charge redistribution between pentamers and Ag(111). Blue (red) areas show electron accumulation (depletion). The white and black dashed lines refer to the position of the P5 assembly and the last Ag layer, respectively. The isosurface level of the plot is set to  ± 13 × 10−3 e Å−3. d Schematic illustration of the charge redistribution at the P5/Ag(111) interface leading to an inward surface dipole (D) moment and a local work function change (\({\phi }_{{{{{\rm{P}}}}}_{5}/{{{\rm{Ag}}}}}\)). The cyclo-P5 layer is colored orange. ΔV* refers to the LCPD change.

Full size image

To better rationalize this, we calculated the charge redistribution at the cyclo-P5/Ag interface (see “Methods”), whose top and side views of isosurfaces of electron accumulation (blue, +13 × 10−3 e Å−3) and depletion (red, −13 × 10−3 e Å−3) are displayed in Fig. 3c. An electron transfer from the Ag(111) substrate to the P atoms of pentamers is observed as a negative charge accumulation located at the cyclo-P5 ring (red). In the P5/Ag gap (marked by white and black dashed lines in the side view in Fig. 3c), charge accumulation/depletion layers emerge below each cyclo-P5 structure, which supports the formation of a hybridized state30,31. We emphasize that such an interface state is not restricted to 2D Xenes on metals since it has been observed for organic/metal systems38.

Between cyclo-P5 rings, we note the absence of in-plane charge redistribution. Considering that the last Ag layer is depleted (red) while each cyclo-P5 has an excess of negative charges (blue), the P5 assembly can be approximated to a lattice of surface dipole moments of D = 1.42 Debye pointing towards the substrate (see arrow in Fig. 3c). The observation of such surface dipole moments is consistent with an increase of the LWF at the P5/Ag interface, fixed to an arbitrary value Δϕ in the diagram of Fig. 3d. While the LWF increase agrees with the LCPD shift to more positive values in force spectroscopy, it is important to note that the LCPD variations are a qualitative indicate of the LWF changes at the atomic scale since it can have a strong distance-dependence on metal substrate39,40. Indeed, the ΔV* cannot directly account for the difference of work function Δϕ = \({\phi }_{{{{{\rm{P}}}}}_{5}/{{{\rm{Ag}}}}}\) - ϕAg shown in Fig. 3d due to the averaging effects of the electrostatic interactions between tip and sample and the uncertainty in the work function of the tip. As will be discussed later, a quantitative experimental value of LWF (0.46 eV) can be obtained by the analysis of IPS spectra.

Finally, the planar cyclo-P5 structure has, in principle, an unpaired electron leading to an anionic state (i.e., cyclo-\({{{{\rm{P}}}}}_{5}^{-}\)), which has been identified by nuclear magnetic resonance (NMR) in the gas phase or as a ligand41,42. Upon adsorption on Ag(111), the cyclo-\({{{{\rm{P}}}}}_{5}^{-}\) anion can coordinate with the Ag atoms below it, leading to a charge redistribution at the interface and the formation of an interfacial state. This charge transfer modifies the amount of charge of the pentamer away from an integer, as confirmed by the Bader charge analysis showing an accumulation of electrons on P atoms (−0.115 e) and an electron depletion (+0.061 e) of the depleted Ag layer. We, therefore, conclude that cyclo-P5 molecules do not retain their anionic character on Ag(111). This conclusion is further corroborated by the absence of a Kondo resonance or spin excitations in dI/dV spectra acquired near the Fermi level43.

Interface state and work function of the P5 assembly

To shed more light on the electronic properties at the P5/Ag interface, we next performed differential conductance measurements (dI/dV) across a P5 domain (see “Methods”). Figure 4a shows the typical dI/dV point-spectra spectra of the network (orange) as compared to Ag(111) (black). We marked with dashed lines the valence band edge maximum (VBE) at −0.5 eV and the conduction band edge minimum (CBE) at 0.3 eV, providing a gap Eg of the P5 assembly to about 0.8 eV. The spectra also show a strong resonance at 2.5 V, which we attribute to tunneling into an interface state (IS). dI/dV maps (Fig. 4b) further reveal the density of states at the valence band at Vs = −0.5 V. This atomic feature evolves to a stripe pattern at Vs = +2.5 V (Fig. 4c, bottom), revealing the spatial modulation of the IS (Fig. 4a) similar to the superstructure shown in STM topographic image of Fig. 2a.

Fig. 4: Tunneling spectroscopy of the P5/Ag interface.
figure 4

a dI/dV point-spectra acquired above the P5 assembly (orange) and on Ag(111) (black), whose locations are shown in the STM inset. (parameters: It = 1 pA, Vs = 500 mV, \({A}_{{{{\rm{mod}}}}}\) = 10 mV, f = 511 Hz). b dI/dV maps at Vs = − 1.25 and 2.5 V, respectively. c STM topographic image of three P5 domains and the corresponding dI/dV maps of the IS modulation. d Scheme of the band alignment and the formation of Stark-shifted IPS (orange lines). e FERT cross-section acquired across the P5 assembly along the dashed line in (a), (Set-points: It = 1 pA, Vs = 500 mV, \({A}_{{{{\rm{mod}}}}}\) = 35 mV, f = 511 Hz). f Single FERT spectra of the P5 assembly and the Ag(111) substrate, showing the series of nth IPS. g Extracted IPS peak voltages as a function of n2/3.

Full size image

To quantify the local change of work function (LWF), we acquired field-effect resonant tunneling (FERT) spectra in order to probe IPS between the cyclo-P5 assembly and silver. Experimentally, FERT spectra (also called dZ/dV spectroscopy) are obtained by sweeping the sample voltage Vs while keeping constant the tunneling current with the STM feedback loop. When the voltage exceeds the local work function, resonant tunneling through the tip-sample junction occurs from the tip into the image potential states, giving rise to a series of peaks in the FERT spectra. From a quasi-classical approximation (Fig. 4d), tunneling resonances occur when the Fermi level of the tip aligns with the Stark-shifted IPS states, see Eq. (1).

$${{{\rm{e}}}}\,V_{{{{\rm{n}}}}}=\phi+{\left(\frac{3n\pi \hslash {{{\rm{e}}}}E}{\sqrt{2m}}\right)\!}^{2/3}$$
(1)

where Vn is the sample voltage for the nth IPS, ϕ is the work function of the sample, m is the free electron mass, and E is the electric field. Interestingly, the comparison of FERT spectra between tip positions enables us to distinguish interfacial charge transfer and quantify the local change of work function at the nanometer scale28,29,30,31.

Figure 4e, f shows the series of IPS states obtained above the cyclo-P5 self-assembly as compared to Ag(111), respectively. The resonance at Vs = 2.2 eV of the orange spectrum of Fig. 4f, which is absent for the Ag one (black), corresponds to the IS. The peaks noted n = 1 to 6 of the black spectra are the IPS states of the pristine Ag substrate. On the P5 assembly, IPS states are clearly shifted to higher voltage when ramping up the electric field (i.e., Vs), which is the signature of the increase of LWF30,31.

A quantitative estimation of the LWF can be obtained from Eq. (1). In Fig. 4g, we plot the voltage position Vn of the IPS states as a function of n2/3 for both the cyclo-P5 network (orange squares) and the Ag(111) substrate (black triangles). By fitting the linear progression of each dataset, we extract the LWF value corresponding to the y-intercepts to ϕAg = 4.49 eV and \({\phi }_{{{{{\rm{P}}}}}_{5}}\) = 4.95 eV, respectively. Considering that our experimental estimate of ϕAg is in agreement with that obtained by ultraviolet photoelectron spectroscopy (UPS)44, we quantify the increase of LWF of Δϕ = 0.46 eV induced by the cyclo-P5 assembly adsorbed on Ag(111). Altogether, the observation of the IS and the shift of IPS resonances in tunneling spectroscopy point to a charge transfer from the Ag substrate to the cyclo-P5 network, leading to the creation of inwards electric dipoles at the interface.

Discussion

In summary, we synthesized phosphorus chains and cyclo-P5 pentamers by depositing phosphorus atoms on atomically flat Ag(111) in an ultra-high vacuum. Using low-temperature AFM with CO-terminated tips, armchair P chains, and cyclo-P5 rings are resolved with atomic precision. DFT calculations support a substantial charge transfer from the Ag substrate to P5 pentamers, which results in a complex charge redistribution at the P5/Ag interface. This further leads to the emergence of an interface state as observed by dI/dV spectroscopy. Using force-voltage spectroscopic measurements, we infer the local increase of work function at the P5 network in comparison to the bare metal substrate and determine the direction of the induced surface dipole moments. We corroborated these measurements with FERT spectroscopy to quantify the LWF increase of 0.46 eV at the P5/Ag interface. Based on our experimental estimates, we summarize in Supplementary Fig. 3 the energy level alignment at the cyclo-P5/Ag interface. Considering that the Ag metal is depleted while the P5 assembly is negatively charged, we conclude that the P5/Ag system behaves as a prototypical p-type semiconductor-metal junction with a Schottky barrier built at the interface. It further shows that this system could have potential applications in field-effect transistors, diodes, or solar cells. Finally, by exploring the fundamental characteristics of the prototypical cyclo-P5/metal interface, our methodology (applicable to other emerging 2D materials and related quantum materials) not only showcases the importance of scanning probe microscopy as a powerful technique to study structural and electronic properties at the atomic scale but also provides insights for improved performances of phosphorus-based devices.

Methods

Sample preparation

The Ag(111) substrate purchased from Mateck GmbH was sputtered by Ar+ ions and annealed at 770 K to eliminate any surface contaminations. Phosphorus atoms were sublimed by heating up a black phosphorus crystal contained in a Knudsen cell in an ultra-high vacuum (UHV). The P flux was estimated using a quartz microbalance. To obtain the phosphorus chains and P5 domains, we annealed the Ag(111) substrate during deposition at temperatures described in ref. 9.

AFM experiments

AFM measurements were performed with commercially available tuning-fork sensors in the qPlus configuration45 equipped with a tungsten tip (f0 = 26 kHz, Q = 10,000–25,000, nominal spring constant k = 1800 N m−1, oscillation amplitude A ≈ 50 pm. Constant-height AFM images were obtained using tips terminated with a single CO in the non-contact frequency-modulated AFM (FMAFM) mode at zero voltage19,46. CO molecules were adsorbed on the sample maintained at below 20 K. Before its functionalization, the apex was sharpened by gentle indentations into the Ag surface. A single CO molecule was carefully attached to the tip following the procedure of ref. 47. Simulations of the AFM images based on the DFT coordinates were carried out using the probe-particle model48. Site-dependent Δf(Z) spectroscopic measurements to determine the atomic buckling of phosphorus pentamers were obtained with CO-terminated tips. The Δf(V) cross-section of 1 × 85 pixels was acquired with Ag-coated metallic tips (tunneling setpoints: It = 1 pA, Vs = 800 mV, Zoffset = +80 pm).

DFT calculations

All DFT calculations were carried out in the Vienna ab initio simulation package (VASP)49 with the projector augmented wave (PAW) method. The generalized gradient approximation (GGA) in the framework of Perdew-Burke-Ernzerhof (PBE)50 was chosen with the plane-wave cutoff energy set at 400 eV for all calculations. The DFT-D351 method of Grimme was employed to describe the van der Waals (vdW) interactions. The geometries of the structures were relaxed until the force on each atom was less than 0.02 eV Å−1, and the energy convergence criterion of 1 × 10−4 eV was met. The Brillouin zone was sampled using Gamma k-mesh with a separation criterion of 0.03. Metal slabs with 3 atomic layers were adopted as the substrate, and the bottom layer was fixed to simulate the bulk. The vacuum spacing between neighboring images was set at least 15 Å along the non-periodic directions to avoid a periodic interaction.